Artificial intelligence at the service of photography
Artificial intelligence (AI) is increasingly becoming part of our daily lives. In the field of photography, it opens the way to a simplified practice so that everyone can intuitively take the photo tool in hand. This technology, still in its infancy, could revolutionize the medium.
The development of AI has been going well in recent years and it has not spared the world of photography. One of the AI technologies, called "deep learning," is based on the way neural cells work. The purpose of this technology is to independently simulate human vision. Although this process is still in its early stages, it is already being implemented in smartphones in particular.
Presentation of the first Smartphone with NPU system by Huawei
The cameras integrated into our everyday companions are almost all based on AI technologies: the program tries to identify the subject of the photo and adapts the shooting settings to obtain a correctly exposed image. The reason why artificial intelligence is more efficient than previous technologies is that it is now possible for it to "learn" to recognize the subject you want to photograph and to adapt according to its "knowledge." This learning method consists of providing a series of images to a learning program that guided by specific instructions will automatically analyze the images, as well as their structures and the way they have been exposed. The result of this learning is a software running on embedded processors (e. g. Huawei, Apple or Samsung NPUs), thus allowing a better adaptation of the camera when shooting.
The major mobile phone manufacturers are all involved in this race, incorporating into their latest models coprocessors that have "learned" to recognize the majority of situations photographed by deep learning algorithms.
Schematic diagram of the separation of the image structure allowing to isolate an element by the Deep Angel algorithm (MIT)
Artificial Intelligence is also of interest to developers of photo and video post-processing software. Researchers at the Massachusetts Institute of Technology (MIT), ETH Zurich and Adobe are working on software that can remove objects from a captured scene (a function called inpainting), for example.
While the removal of elements is not a new technique in itself, it had to be done manually on software programs that provide this possibility, which requires a lot of time and a certain skill.
Deep Angel, the name given to this technology, would make it possible to delete elements of a shot without too much impact on the background. In fact, this algorithm is not yet complete, it creates glitches on the images; as can be seen in the video below.
The primary goal of MIT researchers is not to develop software that perfectly removes one or more elements from a captured scene, but to alert about photo manipulation and its dangers in a world where social networks are now part of the main information channels.
Presentation video of the technology developed by MIT, ETH Zurich and Adobe.
While the integration of intelligent processors seems to be developing in the field of smartphone photography, manufacturers of DSLR and hybrid cameras do not seem to be interested in artificial intelligence technology. Nevertheless, the Start-Up Arsenal has developed a device that can separate elements from the subject and deciphering them. The box, equipped with a dedicated processor and embedded/equipped with a "Deep Learning" algorithm, takes control of the camera to which it is fixed. When triggered, the system with neural networks is able to analyze the scene based on the images "learned" during its processing and a photo database from the Flickr website determines the ideal moment and parameters to take the shot. This assistance is intended to be more efficient than the classic auto mode integrated into digital cameras. The equipment manufactured by Arsenal also offers to connect your smartphone to the camera, allowing you to see the captured subject in real-time and to remotely modify the shooting settings.
Camera controlled remotely thanks to the device developed by the company Arsenal.
Even if artificial intelligence is currently considered as a set of concepts rather than a technology in its own right, it is gradually interfering in our daily lives. It will probably revolutionize the way we approach the photosensitive tool and, above all, make it even more accessible.
However, AI raises questions. What will be the photographer's role once this technique is fully developed? Is the emergence of dull photography, because copied from existing photographs, not to be feared?
In a professional sector that is already undergoing profound transformations linked to new technologies, those questions are arising.